When someone thinks about investing in AI, they think of NVIDIA. But betting everything on a single stock —and the most expensive, most-watched one in the market at that— isn't the only way. The boom needs much more than chips: servers, networking, cooling and, above all, electricity. These are the less obvious layers, and sometimes the less overvalued.
The problem with buying only NVIDIA
NVIDIA is an exceptional company, but concentrating your whole AI bet on it carries two risks: relying on a single company, and paying a price that already assumes years of flawless growth. If something disappoints, the fall can be hard precisely because there was no room for error. Diversifying across the rest of the chain spreads that risk.
The "picks and shovels": who sells to everyone
In a gold rush, selling picks and shovels is often a better business than digging for gold. In AI, the suppliers of chip-making equipment —Applied Materials and Lam Research— win no matter whose chips sell, because every fab needs their machines.
The physical infrastructure of the data center
Chips are only the start. To work, they need a whole infrastructure:
- Vertiv: power and cooling for data centers. The more chips, the more heat to dissipate and power to manage.
- Arista Networks: the ultra-fast networking that connects thousands of chips so they work as one.
- Dell and Super Micro: assemble the servers where the chips live.
The layer almost nobody sees: energy
Here's the most interesting idea. An AI data center consumes as much electricity as a small city, and they're being built at a frantic pace. That has turned utilities into unexpected beneficiaries: suddenly they have a giant, hungry customer. In the U.S., names like Vistra, Constellation Energy and NRG Energy generate the electricity (including nuclear) the new data centers demand. It's a layer far less talked about than the chips and, for that reason, sometimes at more reasonable prices.
But don't assume they're cheap: check it
Beware the opposite trap: thinking that because they're "less famous" these companies are cheap. Some have risen on the euphoria too. The rule doesn't change: look at each one's P/E versus its growth. Type its ticker into the analyzer and compare them sensibly. To see the full map of the ecosystem, go back to the companies behind the AI boom.